''' Modified from: https://github.com/Linear95/CLUB ''' import torch import torch.nn as nn class CLUB(nn.Module): # CLUB: Mutual Information Contrastive Learning Upper Bound ''' This class provides the CLUB estimation to I(X,Y) Method: mi_est() : provides the estimation with input samples loglikeli() : provides the log-likelihood of the approximation q(Y|X) with input samples Arguments: x_dim, y_dim : the dimensions of samples from X, Y respectively hidden_size : the dimension of the hidden layer of the approximation network q(Y|X) x_samples, y_samples : samples from X and Y, having shape [sample_size, x_dim/y_dim] ''' def __init__(self, x_dim, y_dim, hidden_size): super(CLUB, self).__init__() # p_mu outputs mean of q(Y|X) self.p_mu = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim)) # p_logvar outputs log of variance of q(Y|X) self.p_logvar = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim), nn.Tanh()) # self.p_logvar = nn.Sequential(nn.Linear(x_dim, hidden_size//2), # nn.ReLU(), # nn.Linear(hidden_size//2, y_dim)) def get_mu_logvar(self, x_samples): mu = self.p_mu(x_samples) logvar = self.p_logvar(x_samples) return mu, logvar def mi_est(self, x_samples, y_samples): mu, logvar = self.get_mu_logvar(x_samples) # log of conditional probability of positive sample pairs positive = - (mu - y_samples)**2 /2./logvar.exp() prediction_1 = mu.unsqueeze(1) # shape [nsample,1,dim] y_samples_1 = y_samples.unsqueeze(0) # shape [1,nsample,dim] # log of conditional probability of negative sample pairs negative = - ((y_samples_1 - prediction_1)**2).mean(dim=1)/2./logvar.exp() return (positive.sum(dim = -1) - negative.sum(dim = -1)).mean() def loglikeli(self, x_samples, y_samples): # unnormalized loglikelihood mu, logvar = self.get_mu_logvar(x_samples) return (-(mu - y_samples)**2 /logvar.exp()-logvar).sum(dim=1).mean(dim=0) class CLUBSample(nn.Module): # Sampled version of the CLUB estimator def __init__(self, x_dim, y_dim, hidden_size): super(CLUBSample, self).__init__() self.p_mu = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim)) self.p_logvar = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim), nn.Tanh()) def get_mu_logvar(self, x_samples): mu = self.p_mu(x_samples) logvar = self.p_logvar(x_samples) return mu, logvar def loglikeli(self, x_samples, y_samples): mu, logvar = self.get_mu_logvar(x_samples) return (-(mu - y_samples)**2 /logvar.exp()-logvar).sum(dim=1).mean(dim=0) def mi_est(self, x_samples, y_samples): mu, logvar = self.get_mu_logvar(x_samples) sample_size = x_samples.shape[0] #random_index = torch.randint(sample_size, (sample_size,)).long() random_index = torch.randperm(sample_size).long() positive = - (mu - y_samples)**2 / logvar.exp() negative = - (mu - y_samples[random_index])**2 / logvar.exp() upper_bound = (positive.sum(dim = -1) - negative.sum(dim = -1)).mean() return upper_bound/2. class CLUBSample_reshape(nn.Module): # Sampled version of the CLUB estimator def __init__(self, x_dim, y_dim, hidden_size): super(CLUBSample_reshape, self).__init__() self.p_mu = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim)) self.p_logvar = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim), nn.Tanh()) def get_mu_logvar(self, x_samples): mu = self.p_mu(x_samples) logvar = self.p_logvar(x_samples) return mu, logvar def loglikeli(self, x_samples, y_samples): mu, logvar = self.get_mu_logvar(x_samples) mu = mu.reshape(-1, mu.shape[-1]) # (bs, y_dim) -> (bs, 1, y_dim) -> (bs, T, y_dim) -> (bs*T, y_dim) logvar = logvar.reshape(-1, logvar.shape[-1]) y_samples = y_samples.reshape(-1, y_samples.shape[-1]) # (bs, T, y_dim) -> (bs*T, y_dim) return (-(mu - y_samples)**2 /logvar.exp()-logvar).sum(dim=1).mean(dim=0) def mi_est(self, x_samples, y_samples): mu, logvar = self.get_mu_logvar(x_samples) sample_size = mu.shape[0] random_index = torch.randperm(sample_size).long() y_shuffle = y_samples[random_index] mu = mu.reshape(-1, mu.shape[-1]) # (bs, y_dim) -> (bs, 1, y_dim) -> (bs, T, y_dim) -> (bs*T, y_dim) logvar = logvar.reshape(-1, logvar.shape[-1]) y_samples = y_samples.reshape(-1, y_samples.shape[-1]) # (bs, T, y_dim) -> (bs*T, y_dim) y_shuffle = y_shuffle.reshape(-1, y_shuffle.shape[-1]) # (bs, T, y_dim) -> (bs*T, y_dim) positive = - (mu - y_samples)**2 / logvar.exp() negative = - (mu - y_shuffle)**2 / logvar.exp() upper_bound = (positive.sum(dim = -1) - negative.sum(dim = -1)).mean() return upper_bound/2. class CLUBSample_group(nn.Module): # Sampled version of the CLUB estimator def __init__(self, x_dim, y_dim, hidden_size): super(CLUBSample_group, self).__init__() self.p_mu = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim)) self.p_logvar = nn.Sequential(nn.Linear(x_dim, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, hidden_size//2), nn.ReLU(), nn.Linear(hidden_size//2, y_dim), nn.Tanh()) def get_mu_logvar(self, x_samples): mu = self.p_mu(x_samples) logvar = self.p_logvar(x_samples) return mu, logvar def loglikeli(self, x_samples, y_samples): # unnormalized loglikelihood mu, logvar = self.get_mu_logvar(x_samples) # mu/logvar: (bs, y_dim) mu = mu.unsqueeze(1).expand(-1, y_samples.shape[1], -1).reshape(-1, mu.shape[-1]) # (bs, y_dim) -> (bs, 1, y_dim) -> (bs, T, y_dim) -> (bs*T, y_dim) logvar = logvar.unsqueeze(1).expand(-1, y_samples.shape[1], -1).reshape(-1, logvar.shape[-1]) y_samples = y_samples.reshape(-1, y_samples.shape[-1]) # (bs, T, y_dim) -> (bs*T, y_dim) return (-(mu - y_samples)**2 /logvar.exp()-logvar).sum(dim=1).mean(dim=0) / 2 def mi_est(self, x_samples, y_samples): # x_samples: (bs, x_dim); y_samples: (bs, T, y_dim) mu, logvar = self.get_mu_logvar(x_samples) sample_size = x_samples.shape[0] #random_index = torch.randint(sample_size, (sample_size,)).long() random_index = torch.randperm(sample_size).long() # log of conditional probability of positive sample pairs mu_exp1 = mu.unsqueeze(1).expand(-1, y_samples.shape[1], -1) # (bs, y_dim) -> (bs, T, y_dim) # logvar_exp1 = logvar.unqueeze(1).expand(-1, y_samples.shape[1], -1).reshape(-1, logvar.shape[-1]) positive = - ((mu_exp1 - y_samples)**2).mean(dim=1) / logvar.exp() # mean along T negative = - ((mu_exp1 - y_samples[random_index])**2).mean(dim=1) / logvar.exp() # mean along T return (positive.sum(dim = -1) - negative.sum(dim = -1)).mean() / 2